"Should I hire someone for AI, or train the people I already have?" For most companies the answer is to train first. It is rarely a new hire first. Somebody on your payroll is almost certainly using these tools already, and doing it well. Find that person, give them a written scope and protected time, and let what they produce tell you whether a hire is worth making. This article is for the owner of a company with ten to two hundred people who feels pushed toward a new name on the payroll. It sets out the plain answer, the narrow cases where a hire is right, what outside help is for, what a first job description needs, the costs to weigh, and what a lender or buyer will ask about who owns AI in your company. If you have not yet found your internal user, how to spot your AI champion is the place to start, and this article does not repeat it. If the question is whether a tool can take over a role instead of a new hire, can AI replace an employee covers that.
I was COO and then President of a public company that closed more than forty acquisitions, and I ran the integration on most of them. I am a lawyer by training. Now I write fixed-fee AI plans for CEOs of private-equity-backed and owner-led mid-market companies through 013 Labs. This is not legal or HR advice. Anything about hiring, contracts, pay or changes to someone's duties is a question for your own counsel.
The short answer
Train first. A new hire arrives knowing nothing about your customers, your approvals, your jobs or your data, and has to learn all of it before the tools help. Your own people already know the work. Teaching them what the tools can do is a shorter road than teaching a stranger what your business does. Hire only when the work is plainly there, and when it is a real role with real duties that would exist even if the tools disappeared tomorrow.
Why training usually wins
The knowledge that makes these tools useful is knowledge of the work: which quote goes wrong, which report takes a day, which customer email gets asked twenty times a month. That knowledge sits with the people doing the job. A tool is only as helpful as the person who can say what good output looks like, and that is the person who has been doing it.
A hire also arrives with a hope attached. Owners sometimes hope the new person will carry the whole matter so that nobody else has to change. It does not work that way. The tools help where the work is done, in each department, and a single newcomer cannot sit in all of them. If your people have not been taught, the hire becomes a bottleneck, and everyone else keeps waiting.
Training has a smaller downside. If it goes badly, you have spent some hours and learned who is interested. If a hire goes badly, you carry the cost, the disruption and the awkward conversation. How to train employees to use AI covers how to teach people on their own tasks, without a course.
Find who already uses the tools well
Look inside before you look outside. In most companies one or two people got there first. They are not always in a technical job. They are often the person in the office who got tired of a slow task and found a way round it.
You will not find them by asking in a meeting. People worry that admitting they use a tool will get it taken away, or that they will be asked to do more for the same pay. Ask privately and plainly, and make clear that nobody is in trouble. Ask what they have tried, which task it helped with, and what went wrong. Their answers will also tell you quickly whether their use is careful, with the output checked before it goes anywhere, or careless.
Once you have a name, do three things.
Give them a scope in writing. One page. The tasks they may use the tools on, the information they must never put in, and who they check with. A named person with a written scope is a safer position than a company where everyone experiments quietly.
Give them time. A few protected hours a week, taken off their other duties and not added on top. If you ask for the work on top of a full load, it will quietly stop. Think of it as what you would pay for on a hire, spent on someone who already knows the business.
Give them a first job with a result you can check. One repeating task, done a few weeks, measured against your own records. If you cannot say what would count as a result, the scope is not finished.
Whether their role, title or pay should change is a matter for you and your own counsel. Do it deliberately, not by drift. If you do hire, can AI screen resumes covers what a tool may and may not do with the applications.
When a hire does make sense
Sometimes the work is already there and nobody has the capacity or the skills. These are the cases where hiring is reasonable.
A real operations or data role. If the business has needed someone to own its reporting, its processes, its systems or its data for a long time, and the tools now make that job both more valuable and more doable, hire for that role. The duties are real: keeping the numbers clean, owning how work flows between departments, running the systems your staff rely on. The tools are one part of that job, not its whole title.
A first project that has outgrown a part-time owner. If your named person has shown a result and the next steps clearly need someone full time, the role has been proven by the work. That is a far better basis for a job description than a guess.
A gap the current team cannot close. Perhaps nobody can manage your data well enough to feed anything useful. Hire for the data work, described as data work.
What I would not do is hire for "AI person" as a title and nothing else. A job defined only by a technology tends to have no daily duties, no owner and no measure of success. People good enough to be worth hiring can see that, and the ones who cannot are the ones to worry about. Describe the work the person will do on Monday morning, then see whether the tools are part of it.
What outside help is for
Between training your own people and hiring sits outside help, and it has a distinct use. It is for the things your company has no way to learn from the inside quickly.
A second pair of eyes on the plan. Someone who has seen many companies can tell you which tasks are worth starting with, which risks matter in your situation, and which of your hopes are unlikely.
Setting the rules once. A written data rule, an approval process and a clear owner for each tool are things an outsider can help you draft in days, so your internal person starts with a clear frame instead of inventing one.
Borrowing experience while a person grows into the role. Your named person may be good at the tools and new to leading change. A few hours of outside guidance can bridge that.
A view that is not tied to a product. Someone who sells no software has no reason to point you at one. Whichever route you take, who owns AI in a company covers the scope and the three decisions that stay with the owner.
Outside help is not a substitute for ownership. Whoever advises you, a person inside should be named to own the result after they leave. Do I need an AI policy for my small business sets out the short written rules that outside help should leave behind.
If you do hire: what the first job description needs
Write it from the work, not from the technology. A first description for a role like this should have these parts.
- The real duties, in order. What this person will do in the first ninety days, named in terms of your own tasks, reports and systems.
- The systems and data they will touch, and the ones they will not. Be specific about customer, staff and financial information.
- Who they report to and who approves their work. One named person, with the final say written down.
- What they may decide alone and what needs a yes. Buying a tool, changing a process and sharing company information are different decisions.
- How you will know it worked. A measure from your own records, such as time per task, turnaround or corrections, and a date to look at it.
- The checking duty. For anything that reaches a customer, a lender or your books, a named person looks it over before it counts.
- How they will work with the people already doing the job. Part of the role is teaching others, so say so.
Leave out vague lines such as "lead our AI strategy" or "identify opportunities across the business." They describe a hope and not a task. Have your own counsel review the final wording, the terms of employment and anything about confidentiality or ownership of the person's work. I am not able to advise on those.
Costs to weigh, in words
I am not going to give you numbers, because yours depend on your market, your location and your terms. Here is what belongs on each side of the page.
Of a hire: the pay and benefits, the time and cost of finding and interviewing, the weeks before they are useful, the owner's or manager's time spent guiding them, the tools they will want, the risk of a poor fit, and the cost of replacing them if they leave. Add the less visible one: if they leave, the knowledge goes with them unless it was written down.
Of training your own: the hours people spend learning and the work that is not done in those hours, the cost of covering the named person's regular duties while they take on the new scope, the tools and seats you pay for, and the owner's time to set the scope. These costs are real but smaller and easier to reverse.
Of outside help: the fee, set against a fixed, written result, and the time your own people spend with the outside adviser. Ask for the scope and the fee in writing before you start.
Of doing nothing: staff quietly using tools with no rules, with no one checking, and with no way for you to say what they are doing. That also costs something, and it is the one owners tend to leave off. Is AI saving my business money shows how to check what you actually get from the spend, whichever road you choose, and how much does AI cost for a small business helps with the tool costs themselves.
What a lender or a buyer will ask about who owns AI
Whoever looks at your company from outside, a bank, an investor or a buyer, will ask a version of the same question: who is accountable for how AI is used here? They are not testing your technology. They want to know the answer is a person, and that it does not rest on one individual's private habits.
Expect questions such as these.
- Who in the company owns the way these tools are used, and what is their written scope?
- What information is staff allowed to put in, and where is that written down?
- Who checks the output before it reaches a customer, a lender or the accounts?
- If that person left tomorrow, would the company still know what it uses and why?
- What do you pay, what is it for, and what did you cut last quarter?
If your answer is a named internal person with a one-page scope, a short written policy and a record of what you spend, you are in a good position. If your answer is "we hired someone" and nothing is written down, you have created a key-person problem and not solved the question. A buyer will want to see that any gain can be repeated without one individual.
Say only what is true and what you can show. How these arrangements should appear in any formal statement or agreement is for your own counsel and accountant.
Picture a distributor of ninety people whose owner is about to post a job for someone to lead AI. First he asks quietly who already uses these tools, and learns that a customer service lead has been drafting replies for a year and checking each one. He gives her a one-page scope and half a day each week taken off her queue. After a few months her records show faster replies and fewer corrections, and they also show that the real bottleneck is messy order data. He then writes a job description for an operations and data role, built around that work, and asks his own counsel to review the terms. This is a hypothesis, a composite and not a client.
Mistakes that repeat
Hiring before knowing what the job is. Posting a title with no daily duties. Asking staff to take on the scope on top of a full load. Skipping the written scope for the person already doing it. Leaving the knowledge in one head. Expecting outside help to own the result. Telling a lender or buyer more than is written down.
Frequently asked questions
Should I hire someone for AI or train my current staff?
Usually train first. Find the person already using the tools well, give them a written scope and protected time, and measure one task. Hire only if the work has grown beyond them, and then for a real operations or data role.
What if nobody on my team uses AI at all?
Then start with a small, careful trial by one willing person on one repeating task, with a written scope. Outside help can shorten the learning. A hire is still not the first step.
Is it a mistake to hire an AI specialist?
It can be, if the role is defined only by the technology. A specialist hired for real operations or data duties, with a clear scope and a measure of success, is a different matter.
What is outside help actually for?
A second view on your plan, setting written rules once, borrowing experience while someone inside grows into the work, and advice that is not tied to a product. It does not replace a named owner inside the company.
What should a first job description include?
Real duties for the first ninety days, the systems and data in and out of bounds, who approves their work, what they may decide alone, a measure of success, the checking duty and how they will work with current staff. Have your own counsel review it.
Can I change someone's duties or pay because they use AI well?
That is an employment question. Talk to your own counsel before you change anyone's role, title or pay.
What will a lender or buyer ask about who owns AI?
Who is accountable, what staff may put in, who checks the output, whether the company would cope if that person left, and what you spend. A named owner, a written scope and a record of cost answer all five.
Not sure where you stand? Take the 3-minute AI readiness scorecard.